This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

**FUZZY AND STOCHASTIC THEORIES AND THEIR APPLICATIONS IN STRUCTURAL ENGINEERING
**

Z. Kala & A. Omishore, Brno University of Technology, Faculty of Civil Engineering, CZ ABSTRACT

The analysis of the influence of stochastic and fuzzy uncertainties of imperfections on the stochastic and fuzzy uncertainties of the load-carrying capacity of a steel plane frame is presented. On a simple example of a steel plane frame, the analyses of the possibilities of system imperfections implementation as random variables are performed. The influence of uncertainty of random characteristics of the correlation coefficient between the random initial inclinations of the frame columns on the load-carrying capacity of the frame is studied. In the next stage, the system imperfections are considered as fuzzy numbers. The fuzzy load-carrying capacity was determined as a fuzzy number, utilizing the general extension principle. The comparison of fuzzy and stochastic load-carrying capacity with the design standard load-carrying capacity acc. to EUROCODE 3 is performed in the final stage.

1.

INTRODUCTION

Each engineering structure is to a certain degree atypical. The correct estimation of random variables is possible only for exactly determined mass events provided that we have at our disposal an adequate number of experimental results. This is quite accurately fulfilled for geometric and material characteristic of mass-production hot-rolled steel profiles [4, 6]. These apparently credible data are not complete in the event that they are to be used as input random variables in stochastic models of complex, hard to describe and hard to measure systems. Detailed statistical information would require, for e.g. knowledge of correlation matrices, parameters of random fields [2]. Some variables cannot be measured at all (e.g. system imperfections), or the measurement is burdened with high statistical error (e.g. residual stress). The quality of models is, in these cases, given above all by the manner in which the applied methodology compensates with the formalization and effective utilization of uncertainty, which is characteristic in the description of such systems. Let us consider the frame illustrated in Fig. 1 the load-carrying capacity of which is significantly influenced by initial imperfections. The imperfections will be considered as both random variables and fuzzy numbers in the presented article. The random uncertainty is analysed utilizing the stochastic simulation methods Latin Hypercube Sampling (LHS). The fuzzy uncertainty will be analysed utilizing the so-called extension principle. The output variables are

random and fuzzy load-carrying capacity. The application limits and restrictions of mathematical statistics in the formalization of uncertainty are discussed. The fuzzy and stochastic analyses are compared with the load-carrying capacity evaluated according to the EUROCODE 3: stability solution with buckling length and the geometric non-linear solution with initial imperfections.

Figure 1. Geometry of steel plane frame

Figure 2. System imperfection acc. to [9]

2. realizations of eb were within the tolerance limits of + h/100. 6. we shall assume that the mean value me1 = me2 = 0 (perfectly vertical column). to [9] The evaluation of this problem is ambiguous because probability density functions of variables e1 and e2 and their cross-correlation k12 are not known.1 SYSTEM IMPERFECTIONS RANDOM SYSTEM IMPERFECTION In view of the fact that we have inadequate number of statistically exploitable data for system imperfections e1 and e2 at our disposal. 2) and that of random variable eb are found within the tolerance interval of + h/100 (Fig.was necessary. +1〉 by the Monte Carlo (MC) method. 4 and 5. The first file contains the random realization of the pair e1 and e2 with the same sign (shape-wise acc. Figure 4. to Fig.2. Tolerance system limits Multiples of the realizations of variables of the separated files (from the MC solution) in dependence on the correlation k12 for column height of 1 m are illustrated in Fig. Tolerance system limits Figure 3. e2) in such a manner that 95 % of the realizations of valid observations of ea were within the tolerance interval of + h/500 and 95 % of Figure 5. see Fig. the realizations of e1 and e2 with Gaussian distribution . We shall also require that 95% of the realizations of random variable ea are found within the tolerance interval + h/500. These tolerance limits are determined by the tolerance standard [9]. By means of the MC method. Subsequently. The iterative analysis of the problem– repetitive analysis utilizing the MC and the interval-halving methods . 3). After obtaining the standard deviations of the first and the second file. 2) and the second file contains the remaining realizations with opposite sign (shape-wise acc. The histograms obtained for 1 million runs were approximated by continuous probability density functions. Tolerance system limits acc. the union of both files was performed and variables e1 and e2 were statistically evaluated. In practice the stochastic modelling of imperfections is performed in this the following manner. One million random realizations of variables e1 and e2 of the Gaussian distribution of mean value 0. 3). (Fig. This problem can be heuristically analysed as a parametric study (sometimes called the “whatif-study”) for the correlation k12∈〈–1. the standard deviation for both files was designated (multiples of realizations e1. Obtained data file was divided into two. to Fig. standard deviation 1 and correlation k12 were generated by the MC method in the first step.

and 8. to Fig. from tolerance standards or from a small amount of measurements. 2. 7. e2 Figure 8. The behaviour of the standard deviation and skewness of variables e1 and e2 in dependence on the correlation k12 is illustrated in Fig. We generally have only imprecise information. or e1 < 0 and e2 < 0 is multiplied by the value obtained from the broken line in Fig. e. The load-carrying capacity of a frame with imperfections acc. 95 % of realizations of ea and eb are found within the tolerance limits acc. The solution in Fig. i. 3. 2. or e1 < 0 and e2 < 0 is multiplied by the value obtained from the solid line in Fig. standard deviation 1 and with selected correlation k12 is simulated.e. There is inadequate information in [9] for the general solution. for other height values it is necessary to multiply the obtained values with a number that is equal to the column height in meters. 6 and 7 is heuristic. 6.of mean value 0. 6. Standard deviation & skewness of e1. Figure 7. On the contrary. Parameters of system imperfections The problem is relatively easy-to-solve provided that we consider the system imperfections as fuzzy numbers. For symmetrically loaded frames with symmetric boundary conditions it is necessary to consider three basic fuzzy sets of system imperfections. to Fig. The realization of variables e1 > 0 and e2 > 0. This means that the probability density functions (histograms) of the probabilistic variables e1 and e2 and their cross-correlation k12 are unknown. 8). the load-carrying capacity of a frame with imperfections acc. Input random variables cannot be reliably deduced purely from the information provided by the tolerance standards [9]. Imperfections implemented in such a manner satisfy the requirements of the tolerance standard [9]. 3) and away from each other (see Fig.2 FUZZY SYSTEM IMPERFECTION Figure 6. The realization of variables e1 > 0 and e2 > 0. Tolerance system limits acc. 8 is lower than that of a frame without imperfections. to [9] . to Fig.g. the supports of which are determined by the tolerance limits. 3 is higher than that of a frame without imperfections. It is therefore necessary to divide imperfections in different directions into the case of inclination of column top ends (difference in rotation) towards each other (see Fig. The obtained values correspond to a column with height of 1 m. the evaluation of which is burdened with high statistical error. The random realization of system imperfections cannot be obtained in practice from measurements on a higher number of structures. 2 and 3. see Fig.

Membership function acc. The membership function is determined for each set according to the tolerance limits. Figure 12. 3. In each run of the simulation method. the load-carrying capacity was determined with an accuracy of 0. The first author (Kala) of this article is the compiler of the utilized software programme. Figure 10. Membership function acc.Since the frame is symmetric it suffices to consider three sets of system imperfections. The structure was meshed into 11 beam elements (Fig. to [9] The steel frame was solved utilizing the nonlinear Euler incremental method in combination with the Newton-Raphson method. a more general approach is presented upon reaching the allowable strain value [3]. The second criterion for the load-carrying capacity is represented by a loading corresponding to a decrease of the determinant to zero (stability condition). see Fig. the membership functions were considered as triangular. to [9] Figure 11. to [9] . NONLINEAR COMPUTATIONAL MODEL Member geometries may be modelled by means of a beam element with initial curvature in the form of a parabola of the 3rd degree [1]. These criteria present only one variant of the definition of the ultimate limit state. On the contrary it would also be necessary to divide the set A into two sets. This phenomenon occurs at high yield point values with small geometrical member imperfections. Meshing of beam elements Figure 9. Due to the fact that we do not have additional relevant and exploitable information at our disposal. The ultimate one-parametric loading is defined as the lowest value of load-carrying capacity. Geometrical and material nonlinearities were considered. 10. The programme was compiled within the framework of scientific researches and is not intended for commercial use. 11. The first criterion for the load-carrying capacity is given by the loading at which plasticization of the flange is initiated. The load-carrying capacity was evaluated for the basic element material only [7].1 %. Membership function acc. 12). 9. The software programme was compiled in the programming language Pascal.

1. which could significantly dominate over the stochastic uncertainty in complex systems.44884 mm * 6.7314 mm 0.1 LOAD-CARRYING CAPACITY LOAD-CARRYING CAPACITY AS RANDOM VARIABLES Ten random variables were considered in the stochastic analysis. t2 5. Right Column Left Column Figure 14: Design of my own program (in Czech) Figure 13: Hermite density plots Three files of random load-carrying capacity were approximated by the Hermite four parametric (mean. 14. This simple example demonstrates the general drawbacks of present-day probabilistic methods. 14) compiled by the first author (Kala) of this article. e1 Chap. 2. dev.4. 210 GPa 12.6 GPa E * 7. in [5]. 1〉. 1} * Gauss distribution Load-carrying capacity for k12 ∈ {-1. 3.26106 mm * 8. see Tab. 1} is illustrated in Fig. If we do not have at our disposal an adequate number of observations at reproducible conditions.26106 mm * 2.8 MPa 4.3 MPa 16. 210 GPa 12.g.3 MPa 16. The design of the utilized software programme is depicted in Fig. We would obtain results the dispersion of which could be so high that it could completely devalue obtained results. A more detailed mathematical description of fuzzy random functions is contained e. 13 represent the three most important representatives from the set of all results of set k12 ∈ 〈-1. In the event that we do not have at our disposal sufficient valid observations of input random variables. t2 5. deviation * 1.8 MPa 10. 0. Beam t2 9. 4. The programme was compiled in the Delphi interface. 1} * 5. If we do not have accurate information at our disposal. fy 297. Member Symbol Mean value Std. 210 GPa 12. . A possibility of the evaluation of this problem is presented through the utilization of so-called fuzzy-random variables. skewness and kurtosis) Input realizations by LHS methods were evaluated utilizing the programme (Fig.. Young’s Modulus E. 300 runs of the LHS method were performed. yield strength fy (steel S235) and of the imperfections e1. e2 Chap. the model presents the source of vague (fuzzy) uncertainty. distribution utilizing the Statrel programme.6 GPa E * 3. According to sensitivity analysis results [1] the dominant variables are those of the flange thickness t2. e2. 2. std. The correlation k12 is a typical fuzzy characteristic.6 GPa E * 9. Results in Fig. 0. the input data are vaguely (fuzzy) uncertain. fy 297.1 for k12 ∈ {− 1. 0.6601 mm 0. Additional observations are due to economic or technical reasons unavailable. The problem of insufficient valid observations is encountered in a number of engineering applications.6601 mm 0. the variability of natural processes cannot be eliminated by satisfying the requirements of reproducibility by formulating an excess of general conditions.1 for k12 ∈ {− 1. Table 1: Input random variables No. Statistic characteristics were considered according to the experimental results of material and geometric characteristics of structural steel from a dominant Czech manufacturer [4].

consisting of elements z = x•y with a membership function that is given by the minimum of membership values of operators x into fuzzy number Z1 and y into fuzzy number Z2. Let • be an arithmetic operation (e. E = 210 GPa. 11) were considered as input fuzzy numbers. subtraction. 18. division) can then be performed utilizing this function. multiplication. 10. division) and Z1.4. Basic fuzzy arithmetic (addition. Z2 ⊆ R be fuzzy numbers. addition. Other variables were considered as singletons of their characteristic values (fy = 235 MPa. clear and covers all the possible combinations of initial system imperfection implementation. The extension principle in the form of α -cuts was used for the analysis.g. etc. 9. 15. The extension principle then allows the extension of operation • to the operation with fuzzy numbers in the following manner: (1) is a fuzzy number The result of operation Z1 Z2. In the first step only the system imperfections with membership functions (Fig. The fuzzy load-carrying capacities obtained are shown in Figs. Figure 15: Load-carrying capacity – fuzzy set A Figure 18: Load-carrying capacity .2 LOAD-CARRYING CAPACITY AS FUZZY VARIABLES The fuzzy analysis was evaluated according to the general extension principle utilizing the so-called response function.Sets A ∪ B ∪ C . The solution is easy. 17.). 16. Figure 16: Load-carrying capacity – fuzzy set B Figure 17: Load-carrying capacity – fuzzy set C The resulting load-carrying capacity obtained through the union of the three fuzzy sets is depicted in Fig.

13 with the results of fuzzy analysis. emporium in Moscow. the stability solution is more general and provides solution for all types of structures frequently occurring in buildings. The vagueness on the requirements of the standard [8] rests on the fact that it does not strictly designate what combination of initial imperfections to consider. the designer is often forced to design structures of very low weight. to [8] are illustrated in Fig. hall in Humpolec (in Czech Republic) and the winter stadium in Germany. two basic conventional methods: the stability solution with buckling length and the geometric non-linear solution. the imperfections for atypical structures are not listed.4. The fuzzy number of loadcarrying capacity can be evaluated according to the extension principle utilizing the geometric non-linear solution described in Chapter 3. . The uncertainty of the stability solution of more complex systems is given by the uncertainty of the buckling length. STOCHASTIC AND EUROCODE 3 ANALYSES The comparison of the load-carrying capacities given as fuzzy numbers. The basic combinations of member and system imperfections acc. random variables and deterministic according to the standard is difficult. however in the case of higher number of load cases. This approach is sufficiently transparent. whereas the final effects could be very substantial even from the point of view of loss of human life. performing the stability solution for each combination (sometimes of the order of a thousand combinations) is due to the time factor practically unmanageable. 20. fuzzy and deterministic solution EC3 [8] is shown in Fig. The weight of structures designed according to valid standards can be minimized utilizing knowledge on the vagueness of standard regulations and computational models. it has different predicative capabilities. for the analysis of the load-carrying capacity of systems with compressed columns. The problem of the influence of axial forces on the stress state of bars in structural systems is simplified to the estimation of one variable of the buckling length.4 COMPARISON OF FUZZY. This then leads to simplifications in which the buckling length is implemented by one value for all the combinations. The buckling length is a typical vague (fuzzy) characteristic that cannot be measured and statistically evaluated. 4. 1 for this purpose. On the contrary. This limits the practical applicability of the geometric non-linear solution. Although EUROCODE 3 [8] lists the initial geometric imperfections for basic frame types. The fuzzy numbers of input imperfections are defined formatively identical with the Gaussian distribution with parameters from Tab. The core of each fuzzy set is identical to its mean value and the degree of membership of 1 is allocated to it. 19 (imperfections of cross-beam were neglected). see recent collapse of roofs under the weight of snow at the following locations: Exhibition ground in Warsaw. The comparison of the stochastic. which is computationally more accurate but suffers from an inadequacy of input data. Each method is based on a different theory and processes qualitatively different information and hence. Let us compare the results of stochastic analysis in Fig. Furthermore cases when the combination of the aforementioned methods and also simplified procedures according to the 1st order theory can be utilized are listed. The estimation of internal forces and member dimensions by the geometric non-linear solution is limited by the normative estimation of initial geometric imperfections for the solved structure.3 VAGUENESS OF REQUIREMENTS OF EUROCODE 3 EUROCODE 3 [8] lists. It therefore is not a random uncertainty but uncertainty due to the vague requirements of standards and human activity. Figure 19: System and members imperfections Due to economic reasons.

2003. 2. KALA. 4. Heft 1. ISBN 80-02-01770-6. Rome (Italy). CEN. If the design probability of failure is 7. research center project 1M68407700001 and GACzR 103/05/0417.795-808. ISBN 90-5966-040-4. Three density functions advert to the differences due to unfamiliarity of the correlation coefficient k12.. which evidently is not fulfilled. ISSN 0143-974X. U.2E-5 (βd = 3. REFERENCES 1. on Reliability of Structures. 8. 75 (2006). 5.. of. P. The credibility of the stochastic solution is dependent on the validity of input random variables and their correlation. Conf. D. Execution of Steel Structures and Aluminium Structures – Part 2: Technical Requirements for the Execution of Steel Structures. M. 2004. HOFFMANN.. and ROZLÍVKA. S. on Thin-walled Structures. In CD Proc. Sensitivity Analysis of the Stability Problems of Thin-walled Structures. In Proc. MÖLLER. pp. Journal of Constructional Steel Research 61 (2005). ISBN 80-7099-746-X. FAJKUS. 19 were considered in the evaluation of the minimum and maximum of the geometric nonlinear solution.8) then the design load-carrying capacity evaluated according to EN 1990 should be approximately equal to 0. Loughborough (England. All the combinations from Fig. BRDA. 7. J. B-1050 Brussels. 1 would lead to even greater differences among the probability density functions. KOTEŠ. Z. Z. L. ISSN 0143-974X.. 20. REUTER. The quantification of these differences would require a more detailed study inclusive of the random load effects. 20.. to the broadening of the grey region in Fig. BERGMEISTER. Z. VIČAN. . 2005. of Fourth Int. i. Januar 2006. On Structural Safety and Reliability ICOSSAR'05. In Proc. MELCHER. Z. ISSN 0038-9145. Levoča (SR). NOVÁK. BEER.ACKNOWLEDGEMENTS This research was supported by grant KJB201720602 AVČR. CONCLUSIONS The differences between individual solutions are apparent from Fig. Theoretical Basics of Fuzzy Randomness Application to Time Series with Fuzzy Data. 9... TOMKO. J. Sensitivity of Load-carrying Capacity of a Thin-walled Steel Member to the Initial Curvature Shape of its Axis. Probability assessment of system composed of suspension flexural-rigid members. pp. M. J. Bridge Management System of Slovak Railways. to [8] provided that the yield strength was reached in the most stressed section of the frame. ISBN 0 7503 1006-5. UK). of VII Conf. KALA. ENV 1090-2: 2005(E). Journal of Constructional Steel Research 60 (2004)..539-542. EN 1993-1-1:2005(E): Eurocode 3: Design of Steel Structures – Part 1-1: General Rules and Rules for Buildings.1 percentile. Design Characteristics of Structural Steels Based on Statistical Analysis of Metallurgical Products.e. HOLICKÝ. 9th Int. Prague..67-78. A. Conf.835842. pp.415-422. M. 3. The basic probabilistic background of Eurocode is in EN 1990. Stage 34. In Proc. Stahlbau. of the III. KALA. pp.. International Scientific Conference Quality and Reliability of Building Industry. M. pp. K. 2005. We can assume that other variants of correlation implementation among the input variables in Tab. KALA. 5. 6. KMEŤ S.. Differences between design load-carrying capacities evaluated from the geometric non-linear and stability solution of EUROCODE 3 [8] are very alarming. Technologische Eigenschaften von Stählen im europäischen Vergleich. Supplementary observations are often due to technical or economic reasons unavailable.. STRAUSS. Figure 20: Comparison of solutions The design standard values of the load-carrying capacity were evaluated acc. B.

Kale Omishore fuzzy full

Kale Omishore fuzzy full

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd